Converse, Inverse, and Contrapositive

Suppose someone tells you, "If a number is divisible by 4, then it is even." Is it also true that "if a number is even, then it is divisible by 4"? Your instinct should say no — 6 is even but not divisible by 4. That instinct is really a statement about logical structure, not about numbers. Every conditional statement $P \to Q$ has three close relatives. These are obtained by swapping, negating, or swapping-and-negating its two parts: the converse, the inverse, and the contrapositive. The central fact of this article is this: exactly one of these three relatives is guaranteed to share the truth value of the original statement. This holds always, for every possible truth assignment. Knowing which one is the difference between a valid proof strategy and a classic logical fallacy. This matters because mathematics is full of conditional claims. The temptation to silently swap hypothesis and conclusion is a common source of invalid arguments. This happens both among students and in casual reasoning generally ("if it rains, the ground gets wet," concluded backwards to "the ground is wet, so it rained"). Understanding exactly which transformations of a conditional preserve truth, and why, lets you rewrite a statement into a more convenient logically equivalent form. This can be done without changing its meaning. It also explains, at a purely propositional level, why a technique like Proof by Contrapositive is valid. An analogous argument using the converse or inverse is not valid. The big picture is this: $P \to Q$ , its converse $Q \to P$ , its inverse $\lnot P \to \lnot Q$ , and its contrapositive $\lnot Q \to \lnot P$ are four distinct-looking sentences. They are built from the same two atomic pieces $P$ and $Q$ . They are related by a small symmetry group of swap and negate. Truth tables show that this group action splits the four statements into exactly two Logical Equivalence classes: $\{P \to Q,\ \lnot Q \to \lnot P\}$ and $\{Q \to P,\ \lnot P \to \lnot Q\}$ . The converse and the inverse are themselves logically equivalent to each other, being contrapositives of one another. Neither is equivalent to the original conditional in general. This article gives the formal definitions of all four statements. It proves the equivalences and non-equivalences rigorously, via truth tables and via the semantics of the conditional. It works through fully concrete examples with numbers. It also isolates the exact hypothesis-dropping mistake that produces the "affirming the consequent" fallacy. It closes by connecting this material to the proof technique it exists to justify.

Key takeaways

A conditional $P \to Q$ has three variants: converse $Q \to P$ , inverse $\lnot P \to \lnot Q$ , contrapositive $\lnot Q \to \lnot P$ .
The contrapositive is always logically equivalent to the original conditional; this is a theorem, provable by truth table.
The converse and inverse are, in general, not equivalent to the original — a single counterexample statement (with true hypothesis, false conclusion under the swap) kills the equivalence.
The converse and the inverse are logically equivalent to each other, since each is the contrapositive of the other.
Confusing a conditional with its converse is a specific fallacy ("affirming the consequent"); confusing it with its inverse is "denying the antecedent."
Key idea: among the four forms, truth value is determined only by the unordered pair $\{$ negated or not on antecedent side, negated or not on consequent side $\}$ up to the swap-both symmetry — swapping alone breaks equivalence, but swapping-and-negating restores it.

Where this lives today

The contrapositive equivalence is the logical engine behind Proof by Contrapositive, one of the most common indirect proof strategies in analysis, number theory, and topology, used whenever the negated conclusion is easier to work with than the original (e.g. proving "if $n^2$ is even then $n$ is even" by instead showing "if $n$ is odd then $n^2$ is odd"). In computer science, the distinction underlies static program verification and type systems, where confusing a precondition-implies-postcondition contract with its converse is a genuine source of bugs, and it appears explicitly in the design of Hoare-logic proof rules. In digital circuit design, De Morgan-style contraposition identities are used to simplify Boolean expressions and to convert AND/OR gate networks into equivalent NAND/NOR implementations without changing the circuit's truth table. In law and everyday argumentation, the converse/inverse fallacies (affirming the consequent, denying the antecedent) are standard case studies in critical-thinking and philosophy-of-argument courses. In probability and statistics, the same asymmetry appears as the distinction between $P(\text{evidence}\mid \text{hypothesis})$ and $P(\text{hypothesis}\mid\text{evidence})$ — mistaking one for the other is the "prosecutor's fallacy," a continuous-probability cousin of affirming the consequent, resolved quantitatively by Bayes' theorem rather than by simple truth tables, but conceptually the same converse-confusion.

Ask the tutor: Can you show me, step by step, how this equivalence justifies switching to a proof by contrapositive for "if $n^2$ is even, then $n$ is even"?

P

Intuition

Think of $P \to Q$ as a one-way arrow from $P$ to $Q$ : " $P$ is enough to guarantee $Q$ ." The four related statements come from two independent binary choices acting on this arrow: whether you reverse its direction (swap antecedent and consequent) and whether you negate both endpoints. Reversing alone gives the converse; negating alone gives the inverse; doing both gives the contrapositive. The reason the contrapositive survives while the converse and inverse do not is best understood by thinking about what makes a conditional false. The statement $P \to Q$ is false in exactly one situation: $P$ is true and $Q$ is false. Contraposing swaps and negates simultaneously, which turns "P true, Q false" into "not-Q true, not-P false" — precisely the one situation that makes $\lnot Q \to \lnot P$ false too. So the unique falsifying case is preserved by contraposition, and everywhere else both statements are (vacuously or non-vacuously) true. Reversing direction alone, however, moves the falsifying case to a different pair of truth values ( $Q$ true, $P$ false becomes the new failure mode for the converse), which need not coincide with any failure mode of the original. The diagram below places the four statements at the corners of a square, with horizontal edges representing "swap" and vertical edges representing "negate." The diagonals of the square are exactly the equivalence classes: original/contrapositive on one diagonal, converse/inverse on the other.

Figure

The square of opposition for a conditional: diagonals are the equivalence classes

Central theorem

P

P, Q

$P$	$Q$	$P\to Q$
T	T	T
T	F	F
F	T	T
F	F	T

P,Q

$P$	$Q$	$\lnot P$	$\lnot Q$	$P\to Q$	$\lnot Q \to \lnot P$
T	T	F	F	T	T
T	F	F	T	F	F
F	T	T	F	T	T
F	F	T	T	T	T

P, Q

Worked examples

Example 1 (a genuinely asymmetric arithmetic conditional). Let $P$ be " $n$ is divisible by $6$ " and $Q$ be " $n$ is divisible by $3$ ," where $n$ ranges over integers, and fix the specific instance $n = 9$ .

Original: "If $9$ is divisible by $6$ , then $9$ is divisible by $3$ ." Antecedent: $9/6$ is not an integer, so $P$ is false. By the Lemma, a conditional with a false antecedent is true regardless of the consequent, so $P \to Q$ is true here (vacuously).
Converse: "If $9$ is divisible by $3$ , then $9$ is divisible by $6$ ." Here $Q$ (" $9$ divisible by $3$ ") is true ( $9 = 3\cdot 3$ ), and $P$ (" $9$ divisible by $6$ ") is false. By the Lemma, a true antecedent with a false consequent makes the conditional false. So the converse is false at $n=9$ .
Inverse: "If $9$ is not divisible by $6$ , then $9$ is not divisible by $3$ ." Antecedent ( $\lnot P$ ) is true (9 is indeed not divisible by 6); consequent ( $\lnot Q$ ) is false (9 is divisible by 3, so "9 is not divisible by 3" is false). True antecedent, false consequent $\Rightarrow$ the inverse is false at $n=9$ , matching Part 2 of the theorem ( $Q\to P$ and $\lnot P \to \lnot Q$ agree: both false here).
Contrapositive: "If $9$ is not divisible by $3$ , then $9$ is not divisible by $6$ ." Antecedent ( $\lnot Q$ ) is false (9 is divisible by 3). A false antecedent makes the whole conditional true, matching the original's truth value, confirming Part 1 at this instance.

This single instance ( $n=9$ ) already demonstrates the general phenomenon: original and contrapositive agree (both true), converse and inverse agree (both false), but the two pairs disagree with each other. It also shows precisely why "divisible by 3" does not imply "divisible by 6" — the converse direction genuinely fails for a concrete, checkable reason (9 is a multiple of 3 that is not a multiple of 2).

Example 2 (a geometric conditional, checking all four forms symbolically then instantiating). Let $P$ be "quadrilateral $ABCD$ is a square" and $Q$ be "quadrilateral $ABCD$ is a rectangle." Consider the fixed quadrilateral $R$ with vertices $(0,0), (4,0), (4,2), (0,2)$ — a rectangle with side lengths $4$ and $2$ .

Original, $P\to Q$ : "If $R$ is a square, then $R$ is a rectangle." $R$ is not a square (side lengths $4 \ne 2$ ), so $P$ is false, making $P \to Q$ vacuously true.
Converse, $Q \to P$ : "If $R$ is a rectangle, then $R$ is a square." Here $Q$ is true ( $R$ has four right angles and opposite sides equal, satisfying the definition of rectangle), and $P$ is false (as just noted). True antecedent, false consequent, so the converse is false.
Inverse, $\lnot P \to \lnot Q$ : "If $R$ is not a square, then $R$ is not a rectangle." $\lnot P$ is true, $\lnot Q$ is false (since $R$ is a rectangle), so true antecedent/false consequent again: the inverse is false, agreeing with the converse as guaranteed by Part 2.
Contrapositive, $\lnot Q \to \lnot P$ : "If $R$ is not a rectangle, then $R$ is not a square." Antecedent $\lnot Q$ is false ( $R$ is a rectangle). False antecedent $\Rightarrow$ the whole statement is true, agreeing with the original as guaranteed by Part 1.

Notice the mathematical content behind the pattern: "square $\Rightarrow$ rectangle" is a true universally quantified geometric fact (every square satisfies the rectangle definition), so for every quadrilateral, not just $R$ , the original conditional is true and its contrapositive is true. But "rectangle $\Rightarrow$ square" is false as a universal claim, and $R$ is one witness making the converse (and inverse) false. Different rectangles could make the converse true instead (e.g. a $3\times 3$ rectangle, which is a square) — the point of Part 3 of the theorem is only that the converse is not automatically forced to share the truth value of the original, not that it must always differ.

P \to Q

Scope, edge cases & common traps

Situation	What's true	What people wrongly assume
$P \to Q$ is a theorem	Contrapositive $\lnot Q \to \lnot P$ is automatically also a theorem	Converse $Q\to P$ is also automatically a theorem (false in general)
$P \to Q$ and $Q \to P$ both true	Then $P \leftrightarrow Q$ (biconditional) holds	That this always happens whenever $P\to Q$ holds
$P$ false	$P\to Q$ vacuously true, tells you nothing about $Q\to P$	That vacuous truth of $P\to Q$ says anything about the converse's truth value
Proving $P \to Q$	May legally instead prove $\lnot Q \to \lnot P$ (this is Proof by Contrapositive)	May legally instead prove $Q \to P$ or $\lnot P \to \lnot Q$
Natural language "if...then"	Often intends a biconditional colloquially ("if you eat your dinner, [only then] you get dessert")	That mathematical $\to$ carries this implicit converse; it does not, by definition

The mathematician behind it

Origins · 1 minAristotle and how this was discovered

Aristotle

c. 384–322 BCE · Ancient Greek philosophy, Lyceum, Athens

Aristotle's Prior Analytics and Categories laid down the first systematic treatment of conditional and categorical statements, including the recognition that reversing subject and predicate in a claim ("all $S$ are $P$ ") produces a genuinely different claim ("all $P$ are $S$ ") that requires separate justification — the ancestor of the converse/original distinction studied here. The broader apparatus of converse, contrary, and contradictory relationships among quantified statements was systematized into the medieval "square of opposition," a direct conceptual ancestor of the swap/negate square used above for the propositional conditional. The rigorous truth-table treatment of the material conditional itself is a much later, 19th–20th century development (traceable through Boole, Frege, and Post/Wittgenstein's truth tables), but the insight that a statement and its converse are logically independent claims — needing separate proofs — is already fully explicit in Aristotelian syllogistic, where fallacies of "illicit conversion" were catalogued as specific logical errors.

History

- c. 350 BCE — Aristotle's Prior Analytics studies conversion of categorical propositions and identifies invalid conversions as logical errors.

- 13th century — Medieval logicians (e.g. Peter of Spain) formalize the square of opposition, relating a statement to its contrary, converse, and contradictory.

- 1854 — George Boole's The Laws of Thought algebraizes logical connectives, giving the conditional a precise symbolic footing.

- 1880s — Gottlob Frege's Begriffsschrift gives a fully symbolic, quantified treatment of the conditional as a primitive logical connective.

- 1920–1921 — Emil Post and Ludwig Wittgenstein independently introduce truth tables as a systematic decision method for propositional equivalence, making the contrapositive/converse distinction a mechanically checkable fact.

Is the converse ever true, or is it always false?

The converse can absolutely be true — the theorem only says it is not guaranteed to be true just because the original conditional is true. For example, "if $n$ is even, then $n+1$ is odd" has a true converse ("if $n+1$ is odd, then $n$ is even"), and in fact here $P \leftrightarrow Q$ holds. What the theorem rules out is a blanket logical law forcing the converse to inherit the original's truth value for all $P, Q$ . Whether a specific converse happens to be true is a separate mathematical question that must be checked (or proved) on its own merits, statement by statement; it gets no free pass from the truth of the original conditional.

Why does contraposition work but reversing the arrow doesn't?

Because a conditional $P \to Q$ is false in exactly one scenario: $P$ true and $Q$ false. Contraposition (swap and negate simultaneously) sends this scenario to " $\lnot Q$ true and $\lnot P$ false," which is exactly the unique scenario making $\lnot Q \to \lnot P$ false — so the two statements fail together and hold together in all cases, making them equivalent. Reversing the arrow alone (without negating) sends "P true, Q false" to a statement about a different scenario ("Q true, P false") that has no forced relationship to the original falsifying case, so nothing guarantees the truth values line up.

How is the inverse related to the converse, exactly?

The inverse of $P \to Q$ is $\lnot P \to \lnot Q$ , and if you take the contrapositive of the converse $Q \to P$ (swap and negate it), you get $\lnot P \to \lnot Q$ — the same formula as the inverse. So the inverse is literally the contrapositive of the converse. Since contraposition always preserves truth value (the central theorem's Part 1), the inverse and the converse must always share the same truth value as each other, even though neither is guaranteed to share the truth value of the original conditional.

Does this generalize to statements with quantifiers, like "for all

x

P(x) \to Q(x)

Yes, and this is exactly the setting used in the worked examples. Fix any particular object $a$ in the domain; then $P(a)$ and $Q(a)$ are ordinary (unquantified) statements with definite truth values, and everything proved here about $P \to Q$ applies verbatim to $P(a) \to Q(a)$ . The universally quantified claim " $\forall x\, (P(x)\to Q(x))$ " is logically equivalent to " $\forall x\, (\lnot Q(x) \to \lnot P(x))$ " because the equivalence holds pointwise for every $x$ , and a universal quantifier applied to two pointwise-equivalent formulas yields equivalent quantified statements. The quantified converse " $\forall x\,(Q(x)\to P(x))$ " again need not follow, and a single "bad" object $a$ with $Q(a)$ true and $P(a)$ false suffices to refute it — precisely the strategy used with $n=9$ and with the rectangle $R$ .

Is "vacuously true" a loophole that breaks the theorem?

No — vacuous truth is fully accounted for inside the truth table, not an exception to it. When $P$ is false, $P \to Q$ is true regardless of $Q$ (rows 3 and 4 of the Lemma's table), and this fact is used consistently in every direction of every proof above, including the contrapositive equivalence proof. The apparent "loophole" that vacuous truth can mask a false converse (as in Example 2, where $P\to Q$ is vacuously true but $Q \to P$ is false) is exactly the mechanism the counterexample block uses to demonstrate non-equivalence — it is a feature of the proof, not a gap in it.

Why should I care about this if I already know to "prove the contrapositive sometimes"?

Knowing the rule without knowing why it's valid leaves you unable to recognize when a similar-looking move is invalid — for instance, being unable to say why proving the converse instead would not establish the original claim. The truth-table proof here is what lets you certify, with full rigor, that swapping-and-negating is safe while swapping alone is not, rather than trusting a mnemonic. This distinction is also exactly what's tested when a claimed biconditional " $P$ iff $Q$ " is scrutinized: a correct biconditional proof must separately establish both $P\to Q$ and its converse $Q \to P$ , since the theorem guarantees the latter is not automatic.